An Analysis of Municipal Government Expenditures in Japan

The main objective of this study was to examine the determinants of Japanese city expenditures. This study regarded officials of a city as honest agents representing its citizens in the determination of its spending levels and formulated a Japanese model of city expenditures within the neoclassical economic demand framework, paying special attention to two aspects of city officials\u27 fiscal behavior: (1) incremental budgeting and (2) intercity effects on municipal budgeting. Five versions of the Japanese city expenditure model (the basic, dynamic, and three spatially extended models) were specified and estimated with data from more than 100 cities in the Kanto Region for 1970, 1980, and 1990. Relatively modest or high model goodness-of-fit (R$\sp2$) was achieved with the basic and dynamic models (0.73-0.93) for total expenditure in 1970, 1980, and 1990. Additionally, these two models were estimated for ten individual, disaggregated expenditures for these three years. The dynamic model achieved higher R$\sp2$ values in most individual expenditures except for those that mainly consist of capital investments, a finding that strongly supports the hypothesis of city officials\u27 incremental budgeting. Estimation results from the basic and dynamic models also indicate that: (1) income and price elasticities of Japanese cities\u27 total expenditure are close to those reported in past American studies of municipal expenditures; (2) the flypaper effect exists in Japanese city expenditures; and (3) the parameter estimates of crowding for total and individual expenditures compute to be approximately 1.0, just as the past Japanese and American empirical literature has reported. The hypothesis concerning intercity effects on the determination of a city\u27s spending levels was tested by adding a spatially lagged dependent variable to the basic model (MRSA model). The MRSA model failed to attain statistically significant results except for a few cases of welfare expenditure. The spatially autocorrelated error specifications (BSAE and DSAE models) also failed to achieve statistically significant results in most estimation cases; and for some of the statistically significant cases, unexpectedly negative spatial autocorrelation parameter estimates were gained. The magnitudes of all the statistically significant spatial autocorrelation parameter estimates are modest (falling into the interval $\pm$ 0.33)